Vehicle-mounted attitude detection method, system and device

Abstract

The embodiment of the invention provides a vehicle-mounted attitude detection method, system and device, and the method comprises the steps: determining a candidate region of each flow baffle in a plurality of flow baffles based on an initial value of a vehicle-mounted attitude and a vehicle-mounted parameter; based on the candidate area of each flow baffle, determining a candidate point cloud set corresponding to each flow baffle; based on the candidate point cloud set corresponding to each flow baffle, determining a predicted position of the flow baffle; determining target positions of at least two flow baffles in the plurality of flow baffles based on the predicted positions of the plurality of flow baffles; and based on the target positions of the at least two flow baffles, the target value of the vehicle hanging posture is determined.

Description

Technical Field

[0001] This specification relates to the field of unmanned driving, and in particular to a vehicle-hook posture detection method, system, and device. Background Art

[0002] In the future, ports will develop towards full automation. Self-driving container trucks, unmanned terminal equipment, and intelligent dispatching systems will achieve highly coordinated operations, forming a fully automated port operation chain. Unmanned container trucks (unmanned container trucks) are key unmanned equipment in ports, responsible for transporting containers. The trailers of unmanned container trucks undergo complex posture changes (translation, rotation, and tilt) as they move under the traction of the trucks. The autonomous driving system must monitor and adjust the trailer's posture in real time to ensure vehicle stability and safety.

[0003] In addition, ports may be affected by complex factors such as signal interference, occlusion, and noise, which places higher demands on the accuracy of trailer posture detection. Especially for long trailers, slight angle errors may cause large lateral swings at the rear of the vehicle, which may seriously lead to safety accidents.

[0004] Trailer posture detection requires high accuracy, typically within centimeters or within 1 degree of angular error. Finally, trailer posture detection must be performed in a dynamically changing environment, including adapting to varying speeds, accelerations, and unexpected situations. These factors pose significant challenges to high-precision, real-time trailer posture detection.

[0005] Therefore, a vehicle-hook posture detection method, system and device are provided, which can check the vehicle-hook posture in real time and with high precision, thereby improving the safety and efficiency of unmanned container trucks in port operations. Summary of the Invention

[0006] One of the embodiments of the present specification provides a method for detecting a vehicle-hook posture, comprising: determining a candidate area of ​​each of a plurality of baffles based on an initial value of the vehicle-hook posture and vehicle-hook parameters; determining a candidate point cloud set corresponding to each of the baffles based on the candidate area of ​​each of the baffles; determining a predicted position of the baffle based on the candidate point cloud set corresponding to each of the baffles; determining a target position of at least two of the plurality of baffles based on the predicted positions of the plurality of baffles; and determining a target value of the vehicle-hook posture based on the target positions of the at least two baffles.

[0007] One of the embodiments of the present specification provides a vehicle-hook posture detection system, comprising a candidate area determination module configured to determine candidate areas of a plurality of baffles based on an initial value of the vehicle-hook posture and vehicle-hook parameters; a point cloud set determination module configured to determine a plurality of candidate point cloud sets based on the candidate areas of the plurality of baffles; a position identification module configured to determine a predicted position of the corresponding baffle for each candidate point cloud set by performing an identification operation; a position verification module configured to determine a target position of the baffle by performing a verification operation based on the predicted positions of the plurality of baffles; and a posture determination module configured to determine the vehicle-hook posture by performing a posture detection operation based on the target position of the baffle.

[0008] One embodiment of the present specification provides a vehicle-hook posture detection device, comprising a tractor, a vehicle-hook, and at least one processor, wherein the processor is configured to execute the above-mentioned vehicle-hook posture detection method. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, like numbers represent like structures, wherein:

[0010] Figure 1 is a schematic diagram of an application scenario of a vehicle-hook posture detection system according to some embodiments of this specification;

[0011] Figure 2 is a module diagram of a vehicle-hook posture detection system according to some embodiments of this specification;

[0012] Figure 3 is an exemplary flow chart of a vehicle-hook posture detection method according to some embodiments of this specification;

[0013] Figure 4a is an exemplary flow chart of a method for determining a candidate area for a baffle according to some embodiments of this specification;

[0014] Figure 4b is an exemplary schematic diagram of candidate areas according to some embodiments of this specification;

[0015] Figure 5a is an exemplary flow chart of a method for determining a target position of a baffle according to some embodiments of the present specification;

[0016] Figure 5b is a schematic diagram of determining a reference point according to some embodiments of this specification;

[0017] Figure 6is an exemplary flow chart of a method for determining a vehicle-hook posture according to some embodiments of this specification;

[0018] Figure 7 is an exemplary schematic diagram of an evaluation model according to some embodiments of this specification. DETAILED DESCRIPTION

[0019] To more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly describes the drawings required for describing the embodiments. Obviously, the drawings described below are merely examples or embodiments of this specification. Those skilled in the art can apply this specification to other similar scenarios based on these drawings without inventive effort. Unless otherwise apparent from the context or otherwise noted, the same reference numerals in the figures represent the same structure or operation.

[0020] It should be understood that the terms "system," "device," "unit," and / or "module" used herein are a method for distinguishing different components, elements, parts, portions, or assemblies at different levels. However, if other terms can achieve the same purpose, the terms may be replaced by other expressions.

[0021] Unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" do not refer to the singular but include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0022] Flowcharts are used throughout this specification to illustrate the operations performed by systems according to embodiments of this specification. It should be understood that preceding or following operations do not necessarily need to be performed in exact order. Instead, the steps may be processed in reverse order or simultaneously. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0023] Figure 1 It is a schematic diagram of an application scenario of a vehicle-hook posture detection system according to some embodiments of this specification.

[0024] like Figure 1 As shown, the application scenario 100 of the vehicle-hook posture detection system may include a processing device 110 , a network 120 , a terminal 130 , a storage device 140 and a data acquisition device 150 .

[0025] The processing device 110 can process data and / or information obtained from the terminal 130, the storage device 140, and the data acquisition device 150. For example, the processing device 110 can obtain various types of data, such as environmental point cloud data and image data, collected by the data acquisition device 150. For another example, the processing device 110 can process the environmental point cloud data, image data, and other data. In some embodiments, the processing device 110 can analyze and / or process the environmental point cloud data to determine the trailer posture (e.g., deflection angle) of the vehicle (not shown).

[0026] In some embodiments, the processing device 110 can be a single server or a server group. In some embodiments, the processing device 110 can be local or remote. The processing device 110 can be directly connected to the terminal 130, the storage device 140 and the data acquisition device 150 to access the stored or acquired information and / or data. In some embodiments, the processing device 110 can be implemented on a cloud platform. As an example only, the cloud platform can include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-layer cloud, etc. or any combination thereof. In some embodiments, the processing device 110 can be a distributed server group, which can include multiple server nodes.

[0027] The network 120 may include any suitable network that facilitates information and / or data exchange for the application scenario 100. In some embodiments, one or more components of the application scenario 100 (e.g., the terminal 130, the processing device 110, the storage device 140, or the data acquisition device 150) may transmit information and / or data to one or more other components of the application scenario 100 via the network 120. For example, the processing device 110 may obtain environmental point cloud data, image data, etc. from the data acquisition device 150 via the network 120. In some embodiments, the network 120 may be any one or more of a wired network or a wireless network. In some embodiments, the network may be a point-to-point, shared, centralized, or other topological structure, or a combination of multiple topological structures.

[0028] Terminal 130 may include a mobile device 130-1, a tablet computer 130-2, a laptop computer 130-3, or any combination thereof. In some embodiments, terminal 130 may interact with other components in application scenario 100 via network 120. In some embodiments, terminal 130 may receive information and / or instructions input by a user and transmit the received information and / or instructions to processing device 110 via network 120. For example, a user may view the trailer posture of a vehicle (not shown) via terminal 130.

[0029] The storage device 140 can store data and / or instructions. In some embodiments, the storage device 140 can store data obtained from the processing device 110, the terminal 130 and / or the data acquisition device 150. For example, the storage device 140 can store data (such as point cloud data, images) obtained by the data acquisition device 150, etc. In some embodiments, the storage device 140 can store data and / or instructions used by the processing device 110 to execute the exemplary methods described in this specification. For example, the storage device 140 can store instructions for the processing device 110 to execute the methods shown in the various flowcharts. In some embodiments, the storage device 140 may include a large-capacity storage device, a removable storage device, a volatile read-write memory, a read-only memory (ROM), etc., or any combination thereof. In some embodiments, the storage device 140 can be implemented on a cloud platform. In some embodiments, the storage device 140 can be part of the processing device 110.

[0030] The data acquisition device 150 can be used to acquire various types of collected data. For example, the data acquisition device 150 may include a point cloud data acquisition device 150-1 (such as a laser radar), an image acquisition device 150-2 (such as a camera) for acquiring image data, and the like. In some embodiments, the data acquisition device 150 may be mounted on a vehicle (not shown). The data acquisition device 150 can collect various types of collected data from the vehicle's environment. In some embodiments, the vehicle may include an unmanned container truck in a port. The data acquisition device 150 may be mounted on the tractor (such as the roof) of the unmanned container truck, and the data acquisition device 150 can be used to collect environmental data surrounding the unmanned container truck. The environmental data may include environmental point cloud data collected by the point cloud data acquisition device 150-1 and image data collected by the image acquisition device 150-2. In some embodiments, the data acquisition device 150 may transmit the collected data to the processing device 110 and / or the storage device 140 via the network 120 for storage and / or processing. For example, the point cloud data acquisition device 150 - 1 may send the environmental point cloud data to the processing device 110 via the network 120 , so that the processing device 110 determines the target value of the vehicle-hook posture by analyzing the environmental point cloud data.

[0031] The above description is for illustrative purposes only, and actual application scenarios may vary.

[0032] It should be noted that application scenario 100 is provided for illustrative purposes only and is not intended to limit the scope of this application. For those skilled in the art, various modifications or variations can be made based on the description of this specification. For example, application scenario 100 can also include a GPS device to collect vehicle location information, and can also include other sensors such as an inertial measurement unit (IMU) and a wheel speed sensor. However, these changes and modifications do not deviate from the scope of this application.

[0033] Figure 2 It is a module diagram of a vehicle-hook posture detection system according to some embodiments of this specification.

[0034] like Figure 2 As shown, the vehicle-hook posture detection system 200 (hereinafter referred to as the detection system 200 or the detection system) may include a candidate area determination module 210 , a point cloud set determination module 220 , a position identification module 230 , a position verification module 240 and a posture determination module 250 .

[0035] The candidate region determination module 210 is configured to determine a candidate region of each of the plurality of spoilers based on an initial value of the vehicle-hook posture and vehicle-hook parameters.

[0036] In some embodiments, the initial value of the vehicle-hook posture is obtained by at least one of sensor detection, deep learning model detection, and vehicle-hook posture detection module detection.

[0037] In some embodiments, the trailer parameters include at least one of trailer size, a connection point between the trailer and the tractor, a trailer centerline, and spoiler parameters.

[0038] In some embodiments, the baffle parameters are obtained in a manner including at least one of physical measurement and point cloud measurement.

[0039] In some embodiments, the candidate area determination module 210 can be used to determine the range of the initial search area based on the initial value of the vehicle-hook posture and the vehicle-hook parameters; determine the orientation of the initial search area based on the initial value of the vehicle-hook posture; and determine the candidate area of ​​each spoiler based on the range of the initial search area and the orientation of the initial search area.

[0040] In some embodiments, the candidate area determination module 210 can be used to determine the range of the initial search area based on the initial value of the vehicle-hook posture and the vehicle-hook parameters; determine the orientation of the initial search area based on the initial value of the vehicle-hook posture; and determine the candidate area of ​​each spoiler based on the range of the initial search area and the orientation of the initial search area.

[0041] The point cloud set determining module 220 is configured to determine a candidate point cloud set corresponding to each baffle based on the candidate region of each baffle.

[0042] In some embodiments, the point cloud set determining module 220 may be configured to determine a plurality of candidate point cloud sets based on a plurality of candidate regions of spoilers and an initial point cloud set.

[0043] The position identification module 230 is configured to determine a predicted position of the spoiler based on the candidate point cloud set corresponding to each spoiler.

[0044] In some embodiments, the position identification module 230 may also be configured to determine a predicted position of each spoiler based on relative distances between a plurality of point cloud points in its corresponding candidate point cloud set and the centerline of the vehicle trailer.

[0045] The position verification module 240 may be configured to determine target positions of at least two of the plurality of baffles based on the predicted positions of the plurality of baffles.

[0046] In some embodiments, the position verification module 240 may be further configured to determine a deviation based on a positional relationship between the predicted positions of the plurality of baffles and a reference element; and determine target positions of at least two baffles based on the deviation.

[0047] In some embodiments, the position verification module 240 can also be used to determine multiple prediction lines, each prediction line is a line connecting the predicted positions of two different baffles; determine the angle between each prediction line and the reference line; and determine the deviation based on the multiple angles corresponding to the multiple prediction lines, where the deviation is negatively correlated with the number of angles in the multiple angles that meet the consistency requirements.

[0048] In some embodiments, the position verification module 240 may be further configured to determine a set of candidate reference points of each baffle for each baffle; and determine a reference point of the baffle based on the set of candidate reference points of the baffle and a first preset rule.

[0049] In some embodiments, the position verification module 240 can also be used to determine the lateral distance between the reference point of each deflector and the reference point of the vehicle centerline; determine the lateral distance variance value based on multiple reference point lateral distances; and determine the lateral distance variance value as a deviation.

[0050] In some embodiments, the position verification module 240 can also be used to determine multiple baffle sets, each baffle set including at least two baffles; and for each baffle set, based on the positional relationship between the predicted positions of at least two baffles in the baffle set and the reference element, determine the deviation corresponding to the baffle set.

[0051] In some embodiments, the position verification module 240 can also be used to determine multiple sub-areas based on the predicted position of the baffle; for each sub-area, based on the coordinates of each point cloud point in the candidate point cloud set corresponding to the baffle in the reference coordinate system, determine whether the point cloud point is located in the sub-area; in response to the point cloud point being located in the sub-area, determine a candidate point cloud subset corresponding to the sub-area, and the candidate point cloud subset includes the point cloud point; determine a representative point of the candidate point cloud subset through a second preset rule; and determine a candidate reference point set for the baffle based on multiple representative points corresponding to multiple sub-areas.

[0052] In some embodiments, the position verification module 240 can also be used to: for each point cloud point in the candidate point cloud subset, based on the coordinates of the point cloud point in the reference coordinate system, determine whether the lateral distance between the point cloud point and the center line of the vehicle trailer is a maximum value, the lateral distance being the distance in the direction perpendicular to the center line of the vehicle trailer; and in response to the lateral distance between the point cloud point and the center line of the vehicle trailer being a maximum value, determine that the point cloud point is a representative point of the candidate point cloud subset.

[0053] In some embodiments, the position verification module 240 can also be used to: determine the mean of the lateral distances based on multiple lateral distances corresponding to multiple representative points; for each representative point, determine whether the difference between the lateral distance of the representative point and the mean of the lateral distances is greater than a first threshold; in response to the difference between the lateral distance of the representative point and the mean of the lateral distances being not greater than the first threshold, determine the representative point as a candidate reference point.

[0054] In some embodiments, the candidate reference point set of the deflector includes one or more candidate reference points, and the position verification module 240 can also be used to determine the reference point of the deflector based on the candidate reference point set of the deflector and the first preset rule, including: determining the lateral distance between one or more candidate reference points and the center line of the vehicle trailer; sorting one or more lateral distances to determine the maximum lateral distance; and determining the candidate reference point corresponding to the maximum lateral distance as the reference point of the deflector.

[0055] In some embodiments, the position verification module 240 can also be used to sort multiple deviations corresponding to multiple baffle sets to determine the minimum deviation; and use the baffles in the baffle set corresponding to the minimum deviation as at least two baffles; and use the predicted positions of the baffles in the baffle set corresponding to the minimum deviation as the target positions of at least two baffles.

[0056] The posture determination module 250 is configured to determine a target value of the vehicle-hook posture based on the target positions of the at least two spoilers.

[0057] In some embodiments, the posture determination module 250 can be used to determine multiple target reference point sets based on the target position; determine multiple point set combinations based on the multiple target reference point sets, each point set combination including a first target reference point set and a second target reference point set; and determine the vehicle trailer posture based on the point set combination.

[0058] In some embodiments, the posture determination module 250 can be used to determine, for each of the point set combinations, for each candidate reference point in the first target reference point set in the point set combination, a connecting line between the candidate reference point and multiple candidate reference points in the second target reference point set; determine multiple angles between the multiple connecting lines and the reference lines; determine the average of the multiple angles as the candidate value of the vehicle hanging posture; and determine the average of the multiple vehicle hanging posture candidate values ​​corresponding to the multiple point set combinations as the target value of the vehicle hanging posture.

[0059] In some embodiments, the posture determination module 250 may be configured to determine the confidence level of the vehicle-hook posture by taking the variance values ​​of the plurality of vehicle-hook posture candidate values ​​corresponding to the plurality of point sets.

[0060] In some embodiments, the posture determination module 250 may be configured to determine whether the vehicle-hook posture satisfies a first preset condition through a first judgment rule; and perform a vehicle-hook scattered point filtering operation based on the judgment result.

[0061] In some embodiments, determining whether the target value of the vehicle hanging posture meets the first preset condition includes: determining a first judgment result of whether the deviation corresponding to the target position is less than a deviation threshold; determining a second judgment result of whether the confidence of the target value of the vehicle hanging posture is greater than the confidence threshold; and in response to the first judgment result being yes and the second judgment result being yes, determining that the target value of the vehicle hanging posture meets the first preset condition.

[0062] In some embodiments, the posture determination module 250 may be configured to: output a target value of the vehicle hook posture in response to a first preset condition being met; or use an initial value of the vehicle hook posture as the target value of the vehicle hook posture in response to a first preset condition not being met.

[0063] In some embodiments, the posture determination module 250 can be used to determine the confidence level of the vehicle-hook posture using a trained evaluation model, where the evaluation model is a machine learning model.

[0064] It should be noted that the above description of the vehicle-mounted posture detection system 200 and its modules is for convenience of description only and does not limit this specification to the scope of the embodiments cited. It is understandable that for those skilled in the art, after understanding the principle of the system, it is possible to arbitrarily combine the various modules, or form a subsystem to connect with other modules without deviating from this principle. For example, the candidate area determination module 210, the point cloud set determination module 220, the position identification module 230, the position verification module 240 and the posture determination module 250 can be different modules in the system, or one module can realize the functions of two or more of the above modules. For example, the modules can share a storage module, or each module can have its own storage module. Such variations are all within the scope of protection of this specification.

[0065] Figure 3 This is an exemplary flow chart of a vehicle-hook posture detection method according to some embodiments of this specification.

[0066] In some embodiments, process 300 may be performed by detection system 200. Figure 3 As shown, the process 300 includes the following steps.

[0067] Step 310 : determining a candidate region of each of the plurality of baffles based on the initial value of the vehicle-hook posture and the vehicle-hook parameters.

[0068] The trailer is connected to the tractor and moves (such as going straight, turning, etc.) under the power of the tractor. The trailer posture can be used to reflect the position or state of the trailer (i.e., the trailer) at a certain moment in the reference system in which it is located. The reference system can be the three-dimensional space where the tractor and the trailer are located, or the top-down plane of the tractor and the trailer, etc. The detection system 200 can create a three-dimensional coordinate system corresponding to the three-dimensional space, or create a two-dimensional coordinate system corresponding to the top-down plane, and calculate the trailer posture at any moment. For example, the trailer posture can represent the angle or rotation angle of the trailer relative to the tractor at a certain moment T, the inclination angle of the trailer relative to the horizontal plane (such as the ground) where the tractor is located, etc.

[0069] In some embodiments, the trailer posture can be represented by the trailer's angle relative to a reference line. The reference line can be a line connecting the center point of the tractor and the connection point, which can be referred to as the tractor's major axis, the towing axle, or the tractor's centerline. The trailer's angle relative to the reference line is the angle between the trailer's major axis (or the trailer axle, or the trailer's centerline) and the reference line. The connection point refers to the point of contact between the tractor and the trailer when they are mechanically connected via a connection device (such as a towing connection). For example, the connection point can be the center point of the connection device, around which the trailer can rotate.

[0070] When the tractor drives the trailer straight, the trailer centerline is parallel to (or coincides with) the towing axle, and the trailer attitude at this time is 0°; when the tractor turns (such as left or right) or the trailer drifts, the trailer centerline and the towing axle present a certain angle, and the trailer attitude at this time is the angle value (such as 2°, -6°). The positive or negative value of the angle can be used to indicate the direction of deflection (such as the direction corresponding to left or right). For more information about the direction of the angle, see Figure 4b Hereinafter, the trailer posture may be referred to as the trailer angle, trailer deflection angle, or the angle between the trailer and the tractor.

[0071] It should be noted that the trailer posture can be represented by one or more indicators based on actual needs. For example, the trailer posture can also include the aforementioned angle, the lateral distance between the trailer tail (such as the midpoint of the tail) and the baseline, the trailer speed, etc. For another example, the trailer posture can also be used to reflect the changes in the trailer posture over a period of time (such as the trend and amplitude of the angle change).

[0072] The initial value of the vehicle-hook attitude refers to the value of the vehicle-hook attitude when the vehicle-hook attitude detection or calculation is started (such as the current value or the most recently measured value). For example, the initial value can be 0, indicating that the angle between the vehicle-hook and the tractor is 0°.

[0073] The detection system 200 can determine the initial value of the vehicle-hook posture based on multiple different source values, wherein the source values ​​include values ​​determined by multiple different acquisition methods.

[0074] In some embodiments, the initial value of the vehicle-hook posture is obtained by at least one of sensor detection, deep learning model detection, and vehicle-hook posture detection module detection.

[0075] In some embodiments, the source value may include a vehicle-mounted sensor source value obtained by sensor detection, for example, it may be determined by one or a combination of various vehicle-mounted sensors such as an inertial measurement unit (IMU), a GPS, a wheel speed sensor, etc.

[0076] In some embodiments, the source value may include a model source value, which may be obtained based on deep learning model detection. For example, a trained deep learning model may be used to process data collected by sensing devices such as cameras and lidar, and the real-time predicted vehicle-hook posture may be used as the initial value of the vehicle-hook posture.

[0077] In some embodiments, the source value may include a historical source value, which may be obtained based on detection by a vehicle-hook posture detection module (e.g., posture determination module 250 of detection system 200). For example, it may be a historical value of the vehicle-hook posture. A historical value refers to a target value of the vehicle-hook posture calculated by detection system 200 over a period of time. The latest historical value may refer to the historical value most recent to the current time (e.g., the value calculated last time).

[0078] The detection system 200 can select the optimal solution from the source values ​​as the initial value of the vehicle-hook posture based on a preset selection strategy according to the actual situation. For example, the priority order of different source values ​​is historical source value, vehicle-hook sensor source value, and model source value.

[0079] In some embodiments, the selection strategy may include a timeliness strategy, an accuracy strategy, and the like. The timeliness strategy means that, for the above-mentioned multiple source values, the order of their corresponding acquisition time is used as the standard, and the most recently obtained source value can be used as the initial value of the vehicle-hook posture. The accuracy strategy means that, based on the accuracy of the source values ​​under different conditions, the more accurate source value (such as that determined based on prior knowledge) is used as the initial value of the vehicle-hook posture. For example, in rainy and foggy weather, the value predicted in real time by the deep learning model may be subject to greater interference, and its selection priority is reduced.

[0080] In some embodiments of this specification, by providing multiple different initial value sources, the detection system can adapt to different actual situations (such as complex environmental influences, computational load, etc.), thereby achieving greater robustness and redundancy. In addition, if one source value experiences transmission delays or is unreliable, the remaining source values ​​are still available to ensure the normal operation of the module.

[0081] Trailer parameters refer to the physical parameters of the trailer, such as trailer model and weight.

[0082] In some embodiments, the trailer parameters may include at least one of trailer dimensions (eg, length, width, height), a connection point between the trailer and the tractor, a trailer centerline, and spoiler parameters.

[0083] The baffle can be a marker with various shapes. For example, it can be a rectangular parallelepiped, a cylinder, and other shapes.

[0084] In some embodiments, the trailer can be equipped with multiple deflectors. These deflectors can be evenly and symmetrically arranged (e.g., evenly spaced and with the same number on both sides) on both sides of the trailer's body (e.g., the left and right sides when viewed from a bird's-eye view) with a predetermined number of deflectors (e.g., eight). Each deflector is positioned at a predetermined location and protrudes outward by a predetermined distance from the plane of the vehicle body on which it is positioned. Hereinafter, the predetermined protrusion distance may be represented by its width (e.g., 5 cm or 10 cm).

[0085] The baffle parameters may include the number, position distribution, and size (such as length, width, height, etc.) of the baffles.

[0086] In some embodiments, the baffle parameters are obtained in a manner including at least one of physical measurement and point cloud measurement.

[0087] The physical measurement may include measuring with a ruler, for example, measuring the length, width, height, etc. of the baffle based on a ruler.

[0088] Point cloud measurement can be performed by obtaining point cloud data through a sensing device (such as a laser radar, etc.) to determine relevant information of the spoiler (such as size, position on the side of the vehicle trailer, and distribution, etc.).

[0089] The candidate region of the baffle refers to the candidate search region containing the baffle point cloud.

[0090] In some embodiments, the detection system 200 can acquire environmental point cloud data based on detection equipment such as a laser radar. The environmental point cloud data can be point cloud data obtained after the detection equipment detects any object in the environment (e.g., the environment where the tractor and trailer are currently located). The point cloud data can include point cloud data of the tractor, trailer, spoiler, and / or any other object (e.g., an obstacle).

[0091] The candidate regions for the baffles can be represented as the regions covered by the point cloud data of the suspected baffles in the environmental point cloud data. The number of candidate regions can be the same as the number of preset baffles (e.g., 8). The candidate regions can be represented as 3D regions or 2D regions (e.g., the projection of a 3D region on the ground). For more information on determining candidate regions for baffles, see Figure 4a and its description.

[0092] Step 320 : Based on the candidate region of each baffle, determine a candidate point cloud set corresponding to each baffle.

[0093] The candidate point cloud set is the point cloud set obtained after filtering the initial point cloud set. The initial point cloud set is a set of point cloud points corresponding to the environmental point cloud data or a portion thereof. For example, the initial point cloud set may include point cloud data within the coverage area of ​​the environmental point cloud, centered at the connection point, and within a preset coverage radius that includes the tractor and trailer.

[0094] In some embodiments, the detection system 200 may determine multiple candidate point cloud sets based on multiple candidate regions of the baffle and the initial point cloud set. For example, the initial point cloud set may be filtered to select point cloud points located in each candidate region as the candidate point cloud set corresponding to each candidate region.

[0095] Step 330 : Determine the predicted position of each baffle based on the candidate point cloud set corresponding to each baffle.

[0096] The predicted position can be used to indicate a preliminary position of the baffle. The predicted position of the baffle may include a position of a baffle area, an outer edge of the baffle, a center point of the baffle, and the like.

[0097] In some embodiments, the detection system 200 may determine the predicted position of the spoiler based on the relative distances between a plurality of point cloud points in the candidate point cloud set corresponding to the spoiler and the centerline of the vehicle trailer.

[0098] The width of the trailer and the dimensions of each spoiler (e.g., width) are known, and the trailer centerline can be determined based on the initial trailer posture. For any point in the candidate point cloud set, its relative distance to the trailer centerline refers to the perpendicular distance between that point and the trailer centerline or the trailer center plane. The trailer center plane is a plane perpendicular to the ground and passing through the trailer centerline.

[0099] It is understood that the environmental point cloud data or candidate point cloud set is distributed in a three-dimensional environment. The detection system 200 can construct a three-dimensional coordinate system for the vehicle trailer and calculate the distance between each point in the candidate point cloud set and the center plane of the vehicle trailer. Alternatively, the three-dimensional coordinate system can be mapped (or projected) from a bird's-eye view into a two-dimensional coordinate system based on the ground plane, and then the lateral distance between each point in the candidate point cloud set and the center line of the vehicle trailer can be calculated. For ease of description, this distance will be referred to as the first distance below.

[0100] In some embodiments, the detection system 200 may determine the set of point cloud points whose first distance falls within a preset distance threshold as the predicted location of the spoiler. The preset distance threshold may be determined based on the sum of the spoiler width and half the trailer width (e.g., 10 cm). The predicted location in this case may be characterized as the spoiler area.

[0101] In some embodiments, the detection system 200 may determine the set of point cloud points whose first distance equals a preset distance threshold as the predicted position of the deflector. The predicted position in this case may be characterized as the outer edge (surface) of the deflector. The outer edge (surface) refers to the edge (surface) formed by the locations of the voxel points on the deflector with the largest vertical distance from the centerline (surface) of the vehicle trailer.

[0102] It should be noted that the preset distance threshold can be a value obtained by adding a preset extension distance (such as 3 cm) to the width of the deflector, thereby improving the robustness of the predicted position and avoiding excessive deviation in the predicted position caused by errors in the initial value of the vehicle-hook posture and its corresponding vehicle-hook centerline.

[0103] Step 340 : Determine target positions of at least two of the plurality of baffles based on the predicted positions of the plurality of baffles.

[0104] The target position of the baffle refers to the position corresponding to the target baffle used to determine the vehicle-hook posture in step 350. The target position may include the target baffle area, the outer edge of the target baffle, the center point of the target baffle, etc. For example, the target position of the baffle may be the target position corresponding to m (e.g., three) baffles out of n (e.g., four) baffles.

[0105] The target position of the baffle can be used to participate in the vehicle-hook posture detection (or calculation and other analysis processing) to obtain the target value of the vehicle-hook posture. For more information on determining the target position of the baffle, see Figure 5a and its description.

[0106] Step 350 : determining a target value of the vehicle-hook posture based on the target positions of at least two baffles.

[0107] The target value for the vehicle-mounted posture is the final calculated result of the vehicle-mounted posture. For example, the target value may be the current vehicle-mounted posture angle (e.g., 7°). The target value must meet a preset condition. The preset condition may include an accuracy less than a preset angle threshold (e.g., 2°).

[0108] In some embodiments, the detection system 200 may determine whether the target value satisfies a preset condition based on the confidence level of the target value. More information on confidence levels can be found elsewhere in this specification, for example, Figure 6 or Figure 7 .

[0109] In some embodiments, the detection system 200 may select two points on the outer edge from the target positions of at least two baffles, and determine the target value of the vehicle trailer posture based on the angle between the line connecting the two points and the traction axis.

[0110] In some embodiments, the detection system 200 can determine multiple target reference point sets based on the target position, and determine multiple point set combinations based on the multiple target reference point sets, each point set combination including a first target reference point set and a second target reference point set, and then determine the vehicle-hook posture based on the point set combination.

[0111] For more information on determining target values ​​for vehicle-hook posture, see Figure 6 and its description.

[0112] In some embodiments of this specification, by searching for point cloud data of multiple deflectors installed on a trailer and filtering out the point cloud points along the outer edges of the deflectors to calculate the target trailer posture value, noise interference can be eliminated to obtain an accurate trailer posture value. This improves the efficiency and safety of unmanned container trucks in unmanned freight operations at ports.

[0113] Figure 4a is an exemplary flow chart of a method for determining candidate regions of a baffle according to some embodiments of this specification.

[0114] In some embodiments, process 400 may be performed by detection system 200. Figure 4a As shown, process 400 includes the following steps.

[0115] Step 410 : Determine the range of the initial search area based on the initial value of the vehicle-hook posture and the vehicle-hook parameters.

[0116] The initial search area refers to an area including the vehicle-hook point cloud, for example, a rectangular area including the vehicle-hook point cloud.

[0117] In some embodiments, the detection system 200 may determine vertex coordinates of the initial search area in the reference coordinate system based on the vehicle-hook parameters, the connection point, and the first adjustment parameter, and determine a range of the initial search area based on the vertex coordinates of the initial search area.

[0118] The first adjustment parameter can be used to adjust the scope of the initial search area. For example, the first adjustment parameter may include a trailer length extension value and a trailer width extension value. The trailer length extension value and trailer width extension value may be preset extension values, for example, preset ratios of the trailer length and width, respectively, or fixed values.

[0119] In some embodiments, the first adjustment parameter may be related to the initial value of the trailer posture. A larger initial value of the trailer posture indicates a greater lateral distance between the trailer tail and the towing axle, and thus a greater potential error. Consequently, the trailer length extension value and the trailer width extension value may be set to a larger value to provide tolerance for errors in the initial value of the trailer posture, thereby improving robustness.

[0120] In some embodiments, the trailer length extension value and trailer width extension value are also related to environmental conditions. For example, under favorable environmental conditions (such as good weather and no signal interference), the point cloud data is more accurate, and the trailer length extension value and trailer width extension value can be set smaller to reduce the computational complexity. Otherwise, they can be set larger to accommodate the impact of environmental conditions, thereby improving robustness or reliability.

[0121] Figure 4b is an exemplary schematic diagram of candidate areas according to some embodiments of this specification.

[0122] Figure 4b is an exemplary top view in three-dimensional space. Figure 4b As shown, area 417 is the area of ​​the current environment, which can be a rectangular area (or an area of ​​other shapes). The point cloud distributed within the range of area 417 is the environmental point cloud (not shown in the figure), which is obtained by sensing equipment such as laser radar. Area 417 includes a trailer 411 and a tractor 421. The trailer 411 and the tractor 421 are connected at the connection point O. There are multiple baffles on both sides of the trailer (shown in gray rectangles in the figure). Figure 4b As shown, four baffles are arranged on the left and right sides of the vehicle hook 411. The four baffles on each side are arranged side by side at equal intervals. The baffles on the left and right sides can be symmetrical.

[0123] The angle a between the trailer axle 413 of the trailer 411 and the traction axle 422 of the tractor 421 can be used to represent the trailer posture of the trailer 411. For example, when the angle a is 0, it means that the trailer 411 is parallel to the tractor 421. Otherwise, the trailer 411 is deflected by the angle a relative to the tractor 421.

[0124] In some embodiments, the detection system 200 can use the connection point O as the origin of the reference coordinate system, the traction axis 422 as the reference axis (such as the X-axis, not shown in the figure), and the axis perpendicular to the traction axis 422 (such as the Y-axis, not shown in the figure) as another axis to construct a reference coordinate system (such as a two-dimensional coordinate system).

[0125] At a certain time T, angle a can be the initial value of the vehicle hook posture. The detection system 200 can determine the orientation of the vehicle hook shaft 413 based on angle a, and then determine the orientation of the vehicle hook 411 (such as the coordinates of the four vertices of the vehicle hook 411) and the search area 415 based on the vehicle hook parameters (such as the width and length of the vehicle hook). The search area 415 needs to take into account the width of the deflector so that each deflector is included in the search area 415.

[0126] Search area 415 can be directly used as the initial search area. Alternatively, detection system 200 can also expand the length and / or width of search area 415 based on a preset first adjustment parameter by using a trailer length extension value and a trailer width extension value, thereby obtaining initial search area 416.

[0127] Step 420 : Determine the orientation of the initial search area based on the initial value of the vehicle-hook posture.

[0128] The orientation of the initial search area may refer to the direction in which the search area deviates from the towing axis, which may reflect the deflection direction of the vehicle trailer. The orientation of the initial search area may be determined based on an initial value of the vehicle trailer posture, which may be represented in a variety of ways.

[0129] Continue to see Figure 4b , the N direction is parallel to the traction axis 422, and the E direction is opposite to the W direction and perpendicular to the N direction. The orientation of the initial search area can be expressed in the E direction or the W direction. When it deviates to the E direction, the angle a is a positive number, and when it deviates to the W direction, the angle a is a negative number. For example, the initial value of the vehicle hanging posture can be expressed as 5° (or 5°E), then the orientation of the initial search area is the E direction, and the initial value of the vehicle hanging posture can be expressed as -5° (or 5°W), then the orientation of the initial search area is the W direction.

[0130] This is for illustrative purposes only. For example, the direction may be represented by left / right, or clockwise / counterclockwise, etc.

[0131] It should be noted that the orientations of different initial search areas reflect the deflection direction of the vehicle trailer. Taking into account the different deflection directions of the vehicle trailer, there may be detection blind spots on some of the spoilers on both sides of the vehicle trailer. The detection system 200 can select the spoilers on both sides of the vehicle trailer that are consistent with the orientation (such as the spoiler on the right or left side) as the target spoilers for vehicle trailer posture detection.

[0132] Step 430 : determining a candidate region for each baffle based on the range and orientation of the initial search region.

[0133] In some embodiments, the detection system 200 can determine a candidate area for each deflector in a reference coordinate system based on an initial value of the vehicle trailer posture, deflector parameters, and a range and orientation of an initial search area, where the deflector parameters include at least one of a deflector size and a position distribution of the deflector relative to the vehicle trailer.

[0134] Continue to see Figure 4b Detection system 200 can determine candidate regions for each baffle within initial search area 417 based on the parameters of the baffle (the gray rectangular region in the figure). For example, candidate regions corresponding to each baffle (the regions enclosed by the circular dashed lines in the figure) can be determined within initial search area 417 based on the location distribution and width of the baffles. The figure shows candidate region 418 corresponding to only one baffle.

[0135] Figure 4b The shape of the candidate region (such as candidate region 418) is only an example, and its specific shape or size can be consistent with the baffle (such as a rectangular shape). In some embodiments, the candidate region of the baffle can be larger than the size of the baffle.

[0136] It should be noted that the candidate regions of the spoiler can be determined based on a reference coordinate system corresponding to the vehicle hook. For example, a three-dimensional reference coordinate system (not shown) can be constructed with the connection point O as the origin and the vehicle hook axis 413 as the X-axis to obtain the corresponding three-dimensional coordinate values ​​of each candidate region.

[0137] In some embodiments of this specification, a relatively accurate candidate area corresponding to the spoiler can be preliminarily obtained through the initial value of the vehicle hook posture and the vehicle hook parameters and spoiler parameters; at the same time, considering the first adjustment parameter can enhance robustness and provide a reliable basis for further determining the target value of the vehicle hook posture.

[0138] Figure 5a is an exemplary flow chart of a method for determining a target position of a baffle according to some embodiments of the present specification.

[0139] In some embodiments, the process 500 may be performed by the detection system 200. As shown in FIG5 , the process 500 includes the following steps.

[0140] Step 510 : Determine the degree of deviation based on the positional relationship between the predicted positions of the plurality of baffles and the reference element.

[0141] Reference elements can be various preset reference objects. For example, reference elements can include reference lines and reference points. The reference line can be the tractor axle. Since the spoiler is located on both sides of the trailer, the trailer axle can also be used as the reference line.

[0142] The positional relationship between the predicted position and the reference element can be determined based on the type of the reference element. Reference element types may include, but are not limited to, points, lines, and surfaces. Positional relationships can include distances, lines, angles, and so on. For example, the positional relationship may be the vertical distance between the predicted position and the vehicle's axle.

[0143] It should be noted that the positional relationship may also be the relationship between one or more combinations of multiple preset positions and one or more combinations of multiple reference elements. For example, the positional relationship may be the distance between a line connecting two preset positions and a reference line.

[0144] In some embodiments, the reference element includes a reference point. For each baffle, the detection system 200 may determine a set of candidate reference points for the baffle, and determine the reference point of the baffle based on the set of candidate reference points for the baffle and a first preset rule.

[0145] Combine Figure 5b , Figure 5b is a schematic diagram of determining a reference point according to some embodiments of this specification.

[0146] The deflector 511 can be any of the deflectors in the vehicle hanger 411, and the X-axis is the centerline of the vehicle hanger 411 (i.e., the vehicle hanger axis). The candidate region 418 corresponding to the deflector 511 includes a candidate point cloud set consisting of multiple point cloud points (shown as hollow circles, dashed hollow circles, and black solid points in the figure). The candidate region 418 represents the predicted position of the deflector 511. For more information about candidate regions and predicted positions, see Figure 4a and its description.

[0147] In some embodiments, the detection system 200 can determine multiple sub-areas based on the predicted position of the baffle; for each sub-area, based on the coordinates of each point cloud point in the candidate point cloud set corresponding to the baffle in the reference coordinate system, it is determined whether the point cloud point is located in the sub-area, and in response to the point cloud point being located in the sub-area, a candidate point cloud subset corresponding to the sub-area is determined, and the candidate point cloud subset includes the point cloud point.

[0148] Combine Figure 5b ,like Figure 5b As shown, the predicted position of the baffle may correspond to the candidate region 418 of the baffle 511. The detection system 200 may partition the baffle 511 to obtain subregions R1, R2, ..., R10. The partitioning process may be to divide the baffle 511 into equal parts (e.g., 10 equal parts) based on the baffle parameters (e.g., a length of 50 cm) of the baffle 511.

[0149] The detection system 200 can determine whether multiple point cloud points within the candidate region 418 are located in any of the subregions R1, R2, ..., R10 based on their positions (e.g., coordinate values). This allows the detection system 200 to determine whether these points are located in any of the subregions R1, R2, ..., R10. This allows the detection system 200 to obtain a candidate point cloud subset corresponding to each subregion. If a subregion contains no point cloud points, the candidate point cloud subset for that subregion is an empty set.

[0150] like Figure 5b As shown, subregion R1 includes one point cloud point (shown as a solid black circle), subregion R2 includes two point cloud points (shown as a hollow circle and a solid black circle), and subregions R6 and R10 contain no point cloud points. Point cloud points corresponding to the dotted hollow circles in candidate region 418 are located outside any subregion and do not belong to the candidate point cloud subset corresponding to any subregion.

[0151] In some embodiments, the detection system 200 may further select representative points from the subset of candidate point clouds corresponding to the multiple sub-regions as a candidate reference point set for the baffle.

[0152] A representative point is a representative point on each deflector, such as a center point or boundary point. In some embodiments, a representative point can be used to represent the outer boundary surface of the deflector, that is, the point cloud point with the largest vertical distance (or lateral distance) from the centerline of the vehicle trailer relative to other point cloud points in the same candidate area.

[0153] In some embodiments, the representative point may be determined by a second preset rule. The second preset rule refers to a rule or condition for screening the representative point.

[0154] In some embodiments, for each point cloud point in the candidate point cloud subset, the detection system 200 can determine whether the lateral distance between the point cloud point and the vehicle-hook centerline is a maximum value based on the coordinates of the point cloud point in the reference coordinate system. The lateral distance is the distance perpendicular to the vehicle-hook centerline. In response to the lateral distance between the point cloud point and the vehicle-hook centerline being a maximum value, the point cloud point is determined to be a representative point of the candidate point cloud subset. Multiple representative points corresponding to multiple sub-regions can generate a candidate reference point set.

[0155] like Figure 5b As shown, for region R1 and its corresponding candidate point cloud subset, there is only one point cloud point, which is the representative point of the candidate point cloud subset (or the sub-region R1 of the deflector) (shown as a black solid dot); for region R2 and its corresponding candidate point cloud subset, there are two points, of which the point cloud point corresponding to the black solid dot has a greater lateral distance from the vehicle-hook centerline (i.e., the X-axis shown in the figure) than the point cloud point corresponding to the hollow circle. Therefore, the point cloud point corresponding to the black solid dot is the representative point. Similarly, the detection system 200 can determine the representative point of the candidate point cloud subset corresponding to each sub-region.

[0156] In some embodiments, the detection system 200 can determine the mean of the lateral distance based on multiple lateral distances corresponding to multiple representative points; for each representative point, determine whether the difference between the lateral distance of the representative point and the mean of the lateral distance is greater than a first threshold; in response to the difference between the lateral distance of the representative point and the mean of the lateral distance is not greater than the first threshold, determine the representative point as a candidate reference point.

[0157] like Figure 5b As shown, the detection system 200 can calculate the average value of the lateral distances between multiple black solid points and the center line of the vehicle trailer, that is, the lateral distance mean. If the lateral distance corresponding to a black solid point is close to the lateral distance mean, the black solid point will be determined as a candidate reference point.

[0158] The first threshold may be a preset value, which may be used to reflect the degree of proximity between each representative point and the mean of the horizontal distance. The smaller the first threshold is set, the higher the consistency requirement for each representative point is.

[0159] After the candidate reference points are determined, the reference point can be determined based on the first preset rule. Specifically, the detection system 200 can determine the lateral distance between one or more candidate reference points and the center line of the vehicle and trailer, and sort the one or more lateral distances to determine the maximum lateral distance; the candidate reference point corresponding to the maximum lateral distance is determined as the reference point of the spoiler. Figure 5b As shown, the point cloud point 513 in the sub-region R9 has the largest lateral distance from the center line of the vehicle trailer among all the candidate reference points (shown as all the black solid points in the figure). The detection system 200 can determine the point cloud point 513 as the reference point of the deflector 511.

[0160] It should be noted that Figure 5b Only one spoiler of the vehicle trailer 411 is shown. The detection system 200 can determine the reference points corresponding to the other spoilers respectively.

[0161] In some embodiments, the detection system 200 can determine the lateral distance between the reference point of each deflector and the reference point of the vehicle centerline, determine the lateral distance variance value based on multiple reference point lateral distances, and determine the lateral distance variance value as the deviation.

[0162] The lateral distance variance can be used to reflect the consistency of the reference points corresponding to multiple baffles. For example, when the lateral distances of all reference points are relatively close, it indicates a higher degree of consistency among the multiple baffles, and thus reflects a higher accuracy of the selected reference points.

[0163] Deviation reflects the degree of deviation or error between multiple reference points. It's understandable that if the baffle parameters of multiple baffles are consistent, the corresponding reference points (such as those at the outer edges) should lie on the same straight line. However, the real-world environment is more complex, potentially subject to noise and deviations in the initial values ​​of the vehicle-hook posture, which can lead to a certain degree of error or deviation between the multiple reference points.

[0164] In some embodiments of this specification, deviation can be used to filter out inaccurate reference points, thereby eliminating errors in vehicle-trailer posture detection caused by various factors in real-world scenarios (such as point cloud data accuracy, obstacles, noise interference, etc.). This is of great significance in unmanned freight operations. This improves the accuracy of vehicle-trailer posture detection and enhances the safety of unmanned container trucks during missions.

[0165] In some embodiments, the detection system 200 can also determine multiple prediction lines, each prediction line is a line connecting the predicted positions of two different baffles, and determine the angle between each prediction line and the reference line, and then determine the deviation based on the multiple angles, wherein the deviation is negatively correlated with the number of angles that meet the consistency requirements among the multiple angles.

[0166] It should be noted that the predicted position of the baffle can be the area corresponding to the baffle or a portion thereof, and can be represented by one or more point cloud points within the candidate area corresponding to the baffle. In some embodiments, the detection system 200 can filter the preset positions corresponding to the baffle. For example, candidate reference points or reference points can be used as the predicted positions. This reduces computational complexity and avoids unnecessary calculations (e.g., excluding points outside the baffle).

[0167] In some embodiments, meeting the consistency requirement means that multiple angles are within a preset angle range. In some embodiments, meeting the consistency requirement means that the difference values ​​between the multiple angles are within a preset difference threshold range (such as ±2°, etc.). The consistency requirements described in this specification can also be applied to other objects to be measured (such as points, lines, angles, etc.), and the consistency requirements of different objects to be measured can be determined according to the nature or type of the object to be measured. For example, for multiple objects to be measured of point type, the consistency requirement can be evaluated based on whether the lateral distance of each point relative to a reference line (such as the center line of the vehicle hanging) is close (such as less than a threshold); for multiple objects to be measured of line type, the consistency requirement can be evaluated by whether the angle formed by each line and the reference line (such as the center line of the vehicle hanging) is close. The closer it is, the more consistent the multiple objects to be measured are, or the smaller the deviation is.

[0168] For determining the deviation through multiple angles, the detection system 200 can first determine the angle mean of the multiple angles, and based on each angle and the angle mean, determine the angle variance value through a variance algorithm to obtain the deviation. When the number of angles close to the angle mean is greater, the angle variance value is smaller, indicating that the multiple angles are closer, that is, the consistency is higher and the deviation is smaller; when the number of angles close to the angle mean is smaller, the angle variance value is larger, indicating that the difference between the multiple angles is greater, that is, the consistency is worse and the deviation is greater. Therefore, the deviation is negatively correlated with the number of angles that meet the consistency requirements among the multiple angles. The deviation can also be used to measure the screening quality or score of the predicted positions of multiple baffles. The smaller the deviation, the higher the screening quality or the higher the score.

[0169] In some embodiments, the detection system 200 can set a baffle consistency number to ensure the accuracy of vehicle-hook posture detection. The baffle consistency number can be a ratio of baffles or a fixed value, which can be used to indicate the minimum number of baffles required for vehicle-hook posture detection. As an example, for four baffles, the baffle consistency number can be set to 3, indicating that the deviation of the predicted positions corresponding to at least three of the four baffles must be within a preset range. In other words, the lateral distances between the predicted positions of at least three baffles must be relatively close, or that these baffles meet the consistency requirement. While it is not known in advance which baffles meet the consistency requirement, the baffle consistency number must be met.

[0170] In some embodiments, the detection system 200 can determine multiple baffle sets, each baffle set including at least two baffles; and for each baffle set, determine the deviation corresponding to the baffle set based on the positional relationship between the predicted positions of at least two baffles in the baffle set and the reference element.

[0171] The plurality of baffle sets refers to a set consisting of baffles having a specified number of baffle consistency or more according to the consistency requirement.

[0172] Continuing with the example of four baffles (e.g., B1, B2, B3, B4) and a baffle consistency requirement of three, multiple baffle sets include a combination corresponding to the four baffles: {B1, B2, B3, B4}, and the following four combinations corresponding to any three of the four baffles: {B1, B2, B3}, {B1, B2, B4}, {B1, B3, B4}, {B2, B3, B4}, for a total of five baffle sets. Of the multiple (e.g., five) baffle sets described above, at least one baffle set must meet the consistency requirement. The multiple baffles in the baffle set that meet the consistency requirement will be used as computational participants in determining the target value of the vehicle-hook posture.

[0173] In some embodiments, the detection system 200 can determine the positional relationship between the predicted position of each baffle and the reference element according to the method described above for each of the multiple baffles in each baffle set to determine the deviation, thereby obtaining multiple deviations (e.g., 5) corresponding to the multiple baffle sets (e.g., 5). It will be understood that the deviations (i.e., the degree of consistency) corresponding to different baffle sets may vary.

[0174] Step 520 : Determine target positions of at least two baffles based on the deviation.

[0175] In some embodiments, the detection system 200 may determine the target positions of at least two baffles through steps 521 to 523 described below.

[0176] Step 521 : sorting the multiple deviations corresponding to the multiple baffle sets to determine the minimum deviation.

[0177] In some embodiments, the detection system 200 may sort the multiple deviations corresponding to the multiple baffle sets based on a preset sorting algorithm, such as sorting in descending order or ascending order.

[0178] Step 522: The baffles in the baffle set corresponding to the minimum deviation are used as at least two baffles.

[0179] The baffle set with the smallest deviation may be reflected in the plurality of baffle sets, wherein the baffles in the baffle set have the highest consistency.

[0180] In some embodiments, the detection system 200 may sort multiple deviations to determine the minimum deviation. For example, when sorting in ascending order, the deviation with the highest ranking may be used as the minimum deviation, thereby obtaining the baffle set corresponding to the minimum deviation.

[0181] Step 523 : Using the predicted position of the baffle in the baffle set corresponding to the minimum deviation as the target position of at least two baffles.

[0182] In some embodiments, the predicted position of the baffle in the baffle set corresponding to the minimum deviation is the predicted position with the highest consistency, and the detection system 200 can use it as the target position and can be based on Figure 6 The method described is used to further determine the target value of the vehicle-hook posture.

[0183] In some embodiments of this specification, by filtering the preset positions corresponding to individual deflectors, the influence of factors such as environmental noise and obstacles can be eliminated, resulting in a more accurate target position for the deflectors. Furthermore, by introducing a number of consistency requirements, the deviation degree is used to further evaluate the quality of the filtered predicted positions, further ensuring the accuracy of the subsequent target positions, thereby obtaining a more accurate target value for the vehicle trailer posture.

[0184] Figure 6 This is an exemplary flow chart of a method for determining a vehicle-hook posture according to some embodiments of this specification.

[0185] In some embodiments, process 600 may be performed by detection system 200. Figure 6 As shown, process 600 includes the following steps.

[0186] Step 610: Determine a plurality of target reference point sets based on the target position.

[0187] The target reference point set refers to the set of points corresponding to the target locations of the deflectors. For example, for four deflectors on one side of a trailer, the target reference point set may be the set of points corresponding to the target locations of three of the deflectors. In some embodiments, the target reference point set includes target point cloud points corresponding to multiple target locations. For example, it may include points on the outer edges or center points of at least two deflectors.

[0188] Exemplarily, the target reference point set may be {P1, P2, P3}, where P1 represents the target reference point set corresponding to baffle 1 , P2 represents the target reference point set corresponding to baffle 2 , and P3 represents the target reference point set corresponding to baffle 3 .

[0189] Step 620: Determine a plurality of point set combinations based on the plurality of target reference point sets, each point set combination including a first target reference point set and a second target reference point set.

[0190] A point set combination refers to a combination of any two target reference point sets selected from a plurality of target reference point sets. The two target reference point sets in each combination may be referred to as a first target reference point set and a second target reference point set.

[0191] For example, the multiple point set combinations corresponding to the target reference point set {P1, P2, P3} may include the point set combination P 12 、P 13 、P 23 Among them, the point set combination P 12 Represents the combination of the first target reference point set P1 and the second target reference point set P2. The point set combination P 13 Represents the combination of the first target reference point set P1 and the second target reference point set P3. 23 The combination represents the combination of the first target reference point set P2 and the second target reference point set P3.

[0192] Step 630: Determine the vehicle-hook posture based on the point set combination.

[0193] In some embodiments, for each point set combination, the detection system 200 may determine, for each candidate reference point in the first target reference point set, a connecting line between the candidate reference point and multiple candidate reference points in the second target reference point set; determine multiple angles between the multiple connecting lines and the reference lines; determine the average of the multiple angles as the candidate value for the vehicle-hook posture; and determine the average of the multiple candidate values ​​for the vehicle-hook posture corresponding to the multiple point set combinations as the target value for the vehicle-hook posture. The reference line may be a towing axle.

[0194] For example, for the point set combination P 12 , the detection system 200 can select one of the points Poi from the first target reference point set P1 11As candidate reference points, and based on multiple rounds of iterative processing, each round of processing includes selecting multiple points {Poi 21 , Poi 22 ,……,Poi 2n A point Poi in 2m and with Poi 11 Connect to form a straight line L 2m , and thus the straight line L is calculated 2m Angle G with the reference line 2m For example, if the line L 2m Parallel to the reference line, then the angle G 2m is 0°, otherwise it is a non-zero value. After m rounds of iteration, the point set combination P is obtained 12 The corresponding angle set CG 12 {G 21 , G 22 ,……,G 2n The detection system 200 can calculate and record the angle set CG 12 The average value of multiple angles in the point set P 12 Corresponding vehicle-hook posture candidate value Avg 12 According to this principle, the detection system 200 can sequentially combine point sets P 13 and point set combination P 23 Processing is performed to obtain the vehicle-hook posture candidate value Avg 13 and Avg 23 .

[0195] Furthermore, the detection system 200 can calculate the vehicle-hook posture candidate value Avg 12 、Avg 13 and Avg 23 The average value of the vehicle-hook posture is taken as the target value.

[0196] In some embodiments, the detection system 200 may further determine the variance value of multiple vehicle-hook posture candidate values ​​corresponding to the multiple point set combinations as the confidence level of the target value of the vehicle-hook posture.

[0197] The confidence level of the target value of the vehicle hook posture can be used to reflect the reliability of the target value of the vehicle hook posture. For example, when multiple candidate values ​​of the vehicle hook posture are relatively close, it means that the target value of the vehicle hook posture is more reliable.

[0198] The detection system 200 can calculate the variance value of multiple vehicle-hook posture candidate values. The variance value can reflect the difference or disagreement between the multiple vehicle-hook posture candidate values. When the variance value is larger, it means that the difference between them is larger and the confidence is lower. Otherwise, when the variance value is smaller, it means that the difference is smaller and the confidence is higher.

[0199] In some embodiments, the detection system 200 may set a confidence threshold, which may be a preset fixed value (e.g., 0.8). When the confidence level of the target value of the vehicle-hook posture is greater than the confidence threshold, the detection system 200 may use the target value of the vehicle-hook posture as the final target value of the vehicle-hook posture.

[0200] In some embodiments, the confidence threshold may be related to the initial value of the trailer posture. A larger initial value of the trailer posture indicates a greater lateral distance between the trailer tail and the towing axle, and a greater potential error. Therefore, a higher confidence threshold may be set, indicating a higher requirement for the accuracy of the target trailer posture value.

[0201] In some embodiments, the detection system 200 can also determine the confidence level of the target value of the vehicle-hook posture based on the evaluation model. For more information about the evaluation model, see Figure 7 and its description.

[0202] In some embodiments, a first judgment rule is used to determine whether the target value of the vehicle-hook posture satisfies a first preset condition, and based on the judgment result, a vehicle-hook scattered point filtering operation is performed.

[0203] The vehicle-hanging scattered point filtering operation refers to the operation of removing the point cloud points corresponding to the vehicle-hanging in the environmental point cloud data. In some embodiments, in response to the target value of the vehicle-hanging posture meeting the first preset condition, the detection system 200 can perform the vehicle-hanging scattered point filtering operation. When the first preset condition is met, the target value of the vehicle-hanging posture is relatively accurate. At this time, the position of the vehicle-hanging in the environment or the vehicle-hanging range can be accurately obtained, and the vehicle-hanging point cloud points within the range corresponding to the vehicle-hanging are filtered out to prevent the vehicle-hanging point cloud points from being treated as obstacles. It can be understood that when unmanned container trucks are performing unmanned freight tasks, environmental point cloud data can also be used to identify obstacles. Performing the vehicle-hanging scattered point filtering operation can prevent the vehicle from being blocked by scattered points left by the vehicle-hanging, and can also plan the driving path for unmanned container trucks more accurately.

[0204] In some embodiments, the detection system 200 may further perform a scatter point filtering operation based on the target value of the vehicle-hanging posture, vehicle-hanging parameters (e.g., length and width), and a second adjustment parameter. The second adjustment parameter may be used to expand the area corresponding to the vehicle-hanging dimensions and may be the same as or different from the first adjustment parameter. For example, because the target value of the vehicle-hanging posture is relatively accurate, the vehicle-hanging length expansion value and the vehicle-hanging width expansion value of the second adjustment parameter may be smaller than the vehicle-hanging length expansion value and the vehicle-hanging width expansion value of the first adjustment parameter.

[0205] In some embodiments, the first preset condition includes a deviation threshold and a confidence threshold. The detection system 200 can determine a first judgment result of whether the deviation corresponding to the target position is less than the deviation threshold, and determine a second judgment result of whether the confidence level of the vehicle-hook posture is greater than the confidence threshold. In response to a yes result in both the first judgment result and the second judgment result, it is determined that the target value of the vehicle-hook posture satisfies the first preset condition.

[0206] In some embodiments of this specification, the accuracy of the target value of the vehicle-trailer posture can be guaranteed through dual evaluation of the first judgment result and the second judgment result, thereby improving the safety of unmanned container trucks in performing tasks such as unmanned freight transportation.

[0207] In some embodiments, in response to the first preset condition being met, the target value of the vehicle-hook posture is output; in response to the first preset condition not being met, the initial value of the vehicle-hook posture is used as the target value of the vehicle-hook posture.

[0208] In some embodiments, in response to the first preset condition not being met and the initial value of the vehicle-hook posture angle being large, the detection system 200 may also obtain the value of the vehicle-hook posture through sensor detection, deep learning model detection, etc., and select the optimal solution as the target value of the vehicle-hook posture based on the selection strategy. For more information about the selection strategy, see Figure 3 and its description.

[0209] Figure 7 is an exemplary schematic diagram of an evaluation model according to some embodiments of this specification.

[0210] In some embodiments, the detection system 200 can determine the confidence level of the target value of the vehicle-hook posture through a trained evaluation model, where the evaluation model is a machine learning model.

[0211] The evaluation model can be used to determine the confidence level of the target value that can be used to determine the vehicle-hook posture. In some embodiments, the evaluation model can be a trained machine learning model, such as a deep learning model or other customized neural network model.

[0212] In some embodiments, as Figure 7 As shown, the input of the evaluation model 620 may include point cloud data 611 , target value 612 of vehicle-hook posture, vehicle-hook parameters 614 and spoiler parameters 615 , and outputs vehicle-hook posture confidence 630 .

[0213] The point cloud data 611 may be the environmental point cloud data corresponding to the target value 612 of the vehicle-hook posture. Figure 3 and its description.

[0214] The vehicle-hook posture confidence 630 reflects the reliability of the target value 612 of the vehicle-hook posture.

[0215] In some embodiments, as Figure 7 As shown, the input of the evaluation model 620 may also include driving parameters 616 and environmental information 617 .

[0216] Driving parameters 616 may include the driving speed of the trailer (or tractor). Environmental information 617 may include road condition information (such as road width, etc.) and weather information (such as wind speed, wind direction, haze information, etc.).

[0217] In some embodiments, an initial evaluation model can be iteratively trained to obtain a trained evaluation model. Each training sample can include a target value for a sample trailer posture, along with its corresponding sample point cloud data, sample trailer parameters, and sample spoiler parameters. In some embodiments, each training sample can also include sample driving parameters (such as the driving speed of the tractor or trailer) and sample environmental information. Multiple training samples can be obtained based on historical trailer posture detection data. For example, the detection system 200 can generate a set of training samples based on sample point cloud data at a historical time T, sample trailer parameters and sample spoiler parameters corresponding to the sample trailer, sample driving parameters (such as the driving speed at time T), sample environmental information and sample weather information at time T, and the target value for the sample trailer posture calculated by the detection system 200 at the historical time T. The training label can be the confidence level of the sample trailer posture determined by the difference between the target value of the actual trailer posture at the historical time T and the target value of the sample trailer posture at time T. For example, the closer the target value of the sample trailer posture is to the target value of the actual trailer posture, the higher the training label (e.g., 0.95) can be. The training labels can be manually annotated or otherwise annotated. During training, the value of the loss function can be determined based on the difference between the output of the initial evaluation model and the training labels. The parameters of the initial evaluation model can be iteratively updated based on the value of the loss function until the training termination criteria are met (e.g., the loss function converges, a specified number of iterations have been performed, etc.). The updated initial evaluation model can be used as the trained evaluation model.

[0218] In some embodiments of the present specification, through the evaluation model, it is possible to learn the rules of the vehicle trailer posture detected by point cloud data when different vehicle trailers (such as vehicle trailer parameters) are driving in different environments, so that the reliability of the input vehicle trailer posture target value can be automatically determined, thereby providing decision support for the detection system 200 when performing vehicle trailer posture detection.

[0219] It should be noted that the above description of the relevant processes is for illustration and purpose only and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the processes under the guidance of this specification. However, such modifications and changes are still within the scope of this specification.

[0220] While the basic concepts have been described above, it will be apparent to those skilled in the art that the detailed disclosure is merely illustrative and does not limit this specification. Although not explicitly stated herein, various modifications, improvements, and revisions to this specification may be made by those skilled in the art. Such modifications, improvements, and revisions are suggested in this specification and remain within the spirit and scope of the exemplary embodiments of this specification.

[0221] This specification also uses specific terms to describe the embodiments of this specification. For example, "one embodiment," "an embodiment," and / or "some embodiments" refer to a feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "one embodiment," "an embodiment," or "an alternative embodiment" two or more times in different locations in this specification do not necessarily refer to the same embodiment. Furthermore, certain features, structures, or characteristics of one or more embodiments of this specification may be appropriately combined.

[0222] In addition, unless expressly stated in the claims, the order of the processing elements and sequences, the use of alphanumeric characters, or the use of other names described in this specification are not intended to limit the order of the processes and methods of this specification. Although the above disclosure discusses some of the invention embodiments currently considered useful through various examples, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the spirit and scope of the embodiments of this specification. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0223] Similarly, it should be noted that, in order to simplify the presentation of this specification and thus facilitate understanding of one or more embodiments of the invention, the foregoing descriptions of the embodiments of this specification sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this disclosure method does not imply that the subject matter of this specification requires more features than those recited in the claims. In fact, an embodiment may have fewer features than all of the features of a single disclosed embodiment.

[0224] In some embodiments, numbers are used to describe the quantity of components and attributes. It should be understood that such numbers used in the description of the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise stated, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the description and claims are approximate values, which may change according to the required characteristics of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of this specification are approximate values, in specific embodiments, the settings of such numerical values ​​are as accurate as possible within the feasible range.

[0225] Each patent, patent application, patent application publication, and other materials, such as articles, books, specifications, publications, and documents, cited in this specification is hereby incorporated by reference in its entirety. This includes application history documents that are inconsistent with or conflict with the content of this specification, as well as documents (currently or subsequently attached to this specification) that limit the broadest scope of the claims of this specification. It should be noted that if the descriptions, definitions, and / or terminology used in the accompanying materials are inconsistent or conflicting with the content of this specification, the descriptions, definitions, and / or terminology used in this specification will control.

[0226] Finally, it should be understood that the embodiments described in this specification are intended only to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, by way of example and not limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly described and illustrated in this specification.

Claims

1. A method for detecting a vehicle-mounted posture, characterized in that: include: determining a candidate region of each of the plurality of baffles based on an initial value of the vehicle-hook posture and vehicle-hook parameters; Determining a candidate point cloud set corresponding to each of the baffles based on the candidate region of each baffle; Determining a predicted position of the baffle based on a set of candidate point clouds corresponding to each baffle; determining target positions of at least two of the plurality of baffles based on the predicted positions of the plurality of baffles; as well as A target value of the vehicle-hook posture is determined based on the target positions of the at least two baffles.

2. The method according to claim 1, characterized in that Determining the predicted position of the baffle based on the candidate point cloud set corresponding to each baffle includes: For each of the deflectors, a predicted position of the deflector is determined based on relative distances between a plurality of point cloud points in its corresponding candidate point cloud set and the center line of the vehicle trailer.

3. The method according to claim 1, characterized in that Determining the target positions of at least two of the plurality of baffles includes: determining a degree of deviation based on a positional relationship between the predicted positions of the plurality of baffles and a reference element; and Based on the deviation, target positions of the at least two baffles are determined.

4. The method according to claim 3, characterized in that The reference element includes a reference line, and determining the deviation includes: Determining a plurality of prediction lines, each prediction line being a line connecting two different predicted positions of the baffle; determining an angle between each of the prediction lines and the reference line; The degree of deviation is determined based on the angle, and the degree of deviation is negatively correlated with the number of angles that meet the consistency requirement among the multiple angles.

5. The method according to claim 3, characterized in that The reference element includes a reference point, and the reference point is determined based on the following method: For each of the baffles, determining a set of candidate reference points for the baffle; and The reference point of the baffle is determined based on the candidate reference point set of the baffle and a first preset rule.

6. The method according to claim 5, characterized in that Determining the deviation includes: Determining a lateral distance between a reference point of each of the baffles and a reference point of the vehicle trailer centerline; Determining a lateral distance variance value based on the plurality of reference point lateral distances; and The lateral distance variance value is determined as the deviation.

7. The method according to claim 3, characterized in that Determining the deviation includes: determining a plurality of baffle sets, each of the baffle sets comprising at least two baffles; and For each baffle set, a deviation corresponding to the baffle set is determined based on a positional relationship between the predicted positions of at least two baffles in the baffle set and the reference element.

8. The method according to claim 1, characterized in that Determining a target value of the vehicle-hook posture based on the target positions of the at least two baffles includes: Based on the target position, determining a plurality of target reference point sets; determining a plurality of point set combinations based on the plurality of target reference point sets, each point set combination comprising a first target reference point set and a second target reference point set; and The vehicle-hook posture is determined based on the point set combination.

9. The method according to claim 1, characterized in that The determining of a candidate region for each of the plurality of baffles based on the initial value of the vehicle-hook posture and the vehicle-hook parameters comprises: Determining a range of an initial search area based on the initial value of the vehicle-hook posture and the vehicle-hook parameters; Determining the orientation of the initial search area based on the initial value of the vehicle-hook posture; and A candidate region for each of the baffles is determined based on a range of the initial search region and a direction of the initial search region.

10. The method according to claim 1, characterized in that The method further comprises: Determining, by a first judgment rule, whether the target value of the vehicle-hook posture satisfies a first preset condition; and Based on the judgment result, a vehicle-mounted scattered point filtering operation is performed.

11. The method according to claim 1, wherein The method further comprises: The confidence level of the vehicle-hook posture is determined by a trained evaluation model, where the evaluation model is a machine learning model.

12. A vehicle-hook posture detection system, characterized in that: include: a candidate region determination module configured to determine a candidate region for each of the plurality of baffles based on an initial value of the vehicle-hook posture and vehicle-hook parameters; a point cloud set determining module, configured to determine a candidate point cloud set corresponding to each baffle based on the candidate area of ​​each baffle; a position identification module configured to determine a predicted position of the baffle based on a set of candidate point clouds corresponding to each baffle; a position verification module configured to determine a target position of at least two of the plurality of baffles based on the predicted positions of the plurality of baffles; as well as The posture determination module is configured to determine a target value of the vehicle-hook posture based on the target positions of the at least two baffles.

13. A vehicle-hook posture detection device, characterized in that: The invention comprises a tractor, a trailer and at least one processor, wherein the processor is configured to execute the method according to any one of claims 1 to 11.

Images